Tunable Solvation Effects on the Size-Selective Fractionation of Metal

Gregory Von White , II , Matthew Grant Provost , and Christopher Lawrence Kitchens ... Steven R. Saunders , Mario R. Eden , and Christopher B. Roberts...
0 downloads 0 Views 250KB Size
22852

J. Phys. Chem. B 2005, 109, 22852-22859

Tunable Solvation Effects on the Size-Selective Fractionation of Metal Nanoparticles in CO2 Gas-Expanded Solvents Madhu Anand, M. Chandler McLeod, Philip W. Bell, and Christopher B. Roberts* Department of Chemical Engineering, Auburn UniVersity, Auburn, Alabama 36849 ReceiVed: August 19, 2005; In Final Form: October 5, 2005

This paper presents an environmentally friendly, inexpensive, rapid, and efficient process for size-selective fractionation of polydisperse metal nanoparticle dispersions into multiple narrow size populations. The dispersibility of ligand-stabilized silver and gold nanoparticles is controlled by altering the ligand tailssolvent interaction (solvation) by the addition of carbon dioxide (CO2) gas as an antisolvent, thereby tailoring the bulk solvent strength. This is accomplished by adjusting the CO2 pressure over the liquid, resulting in a simple means to tune the nanoparticle precipitation by size. This study also details the influence of various factors on the size-separation process, such as the types of metal, ligand, and solvent, as well as the use of recursive fractionation and the time allowed for settling during each fractionation step. The pressure range required for the precipitation process is the same for both the silver and gold particles capped with dodecanethiol ligands. A change in ligand or solvent length has an effect on the interaction between the solvent and the ligand tails and therefore the pressure range required for precipitation. Stronger interactions between solvent and ligand tails require greater CO2 pressure to precipitate the particles. Temperature is another variable that impacts the dispersibility of the nanoparticles through changes in the density and the mole fraction of CO2 in the gas-expanded liquids. Recursive fractionation for a given system within a particular pressure range (solvent strength) further reduces the polydispersity of the fraction obtained within that pressure range. Specifically, this work utilizes the highly tunable solvent properties of organic/CO2 solvent mixtures to selectively size-separate dispersions of polydisperse nanoparticles (2 to 12 nm) into more monodisperse fractions ((2 nm). In addition to providing efficient separation of the particles, this process also allows all of the solvent and antisolvent to be recovered, thereby rendering it a green solvent process.

Introduction Nanoparticle-based technologies take advantage of the fact that materials built from particles less than a critical length1 display unique chemical and physical properties. These properties depend heavily on the size, shape, and composition of the nanoparticles. The size-dependent properties of nanoparticles allows one to engineer them to have a specific function such as in catalysts,2-4 quantum dots for optical properties,5,6 and medical applications.7 The preparation of monodisperse metal particles is also necessary to study the effects of size on their novel applications including size-dependent conduction of electrons in Ag nanoparticles8 and size-dependent oxidation with Au catalysts.9,10 Moreover, monodisperse nanoparticles are also critical in the production of high-quality ordered arrays and ordered thin films.11-14 There are numerous methods to produce metal nanoparticles, including simple solution-based techniques such as reverse micelle synthesis15,16 and two-phase arrested precipitation methods.17 While these particular solution-based techniques are attractive due to their simplicity, they often result in the synthesis of particle sizes with a wide size range (e.g., 2 to 12 nm). As such, post-synthesis processing is required to further refine the size distribution to the desired narrow monodisperse range. Herein we will use the relaxed definition of monodisperse particles as being samples that have standard deviation, σ , of * Address correspondence to this author. E-mail: [email protected]. Phone: (334) 844-2036. Fax: (334) 844-2063.

diameter less than 5%18 to 10%. A variety of post-synthesis techniques have been developed to narrow size distributions including the use of liquid antisolvents14,19,20 to selectively control precipitation, isoelectric focusing electrophoresis (IEF),21 and chromatography techniques.22 As an example, Sigman et al.20 used ethanol as an antisolvent and centrifugation to sizeselectively precipitate and separate a polydisperse dispersion of silver nanoparticles capped with dodecanethiol ligands into monodisperse particle fractions. In these liquid antisolvent nanoparticle precipitation techniques, ligand-capped particles are first dispersed in solution where the interaction between the solvent and the ligand tails provides enough repulsive force to overcome the inherent van der Waals attraction between the particles that would otherwise result in agglomeration and precipitation. Through the addition of an antisolvent, the resultant poorer solvent mixture interacts less with the ligand tails than did the pure solvent, thereby reducing the ability of the solvent/antisolvent mixture to disperse the particles. Larger particles possess greater interparticle van der Waals attractions and therefore precipitate first upon worsening solvent conditions followed by subsequent precipitation of the smaller sized particles with further addition of antisolvent. Applying centrifugation then provides an external force to accelerate the precipitation process. Repetition of this antisolvent/centrifugation method on the separated particles can result in narrow particle size distributions, σ < 5%; however, the whole process is both solvent and time intensive. It is also difficult to obtain an a priori desired particle size through this

10.1021/jp0547008 CCC: $30.25 © 2005 American Chemical Society Published on Web 11/12/2005

Size-Selective Fractionation of Metal Nanoparticles separation process in a repeatable manner simply by changing the composition of the liquid antisolvent/solvent pair. To overcome this limitation, and to provide improved control over size-selective precipitation, McLeod et al.23 developed an antisolvent precipitation technique based on the pressure tunable solvent properties of gas-expanded liquid mixtures (liquid solvents pressurized with CO2) as described later in this paper. Research in the area of nanoparticle processing using compressed and supercritical fluid solvents has shown that the pressure and temperature tunable solvent properties in these systems provide a means to control the size of nanoparticles that can be synthesized and/or dispersed.24-27 Shah et al. demonstrated the size-selective dispersion of dodecanethiol coated nanoparticles in supercritical ethane by density tuning.28 They illustrated that with the change in solvent density, the dispersable particle size could be adjusted where the largest particle sizes were dispersed at the highest pressure. However, ethane is a feeble solvent and very high pressures of around 414 bar were required to disperse particles of only 3.7 nm in size. Efficient solvent-based separation techniques for a wide range of nanoparticle sizes would require better solvent strength than these supercritical solvents are able to provide at significantly lower pressures. Gas-expanded liquid systems, on the other hand, provide a wide range of solvent properties (from liquid-like to gas-like) that are widely tunable with simple adjustments in gas pressure thereby providing further opportunity for nanoparticle precipitation and separation.23 For example, Han and co-workers recently precipitated nanoparticles from AOT reverse micelles in liquid isooctane using pressurized CO2 as an antisolvent.29-31 This antisolvent effect with CO2 is available because compressed gases such as CO2 dissolve into organic liquids and expand the liquid volume significantly while also altering the liquid’s solvation characteristics. These pressurized liquid solutions of organic solvent and CO2 mixtures are commonly referred to as gas-expanded liquids (GELs). Among their many applications, GELs have been used as tunable reaction media,32-34 as adjustable solvents for separations,35-38 in the switching of fluorous compound solubilities,39 and in gas antisolvent (GAS) precipitation techniques for organic and polymer microparticle formation.40 CO2 is an excellent choice in these gas-expanded liquids as it is a very weak solvent even at high pressures41 and has no dipole moment and very low refractive index.24 As such, the solvent strength of CO2-expanded organic liquid solutions can be varied from that of the pure organic to that of liquid CO2 at pressures below the vapor pressure of CO2. McLeod et al.23 utilized the highly tunable solvent properties of CO2-expanded organic solvents to size-selectively precipitate and separate ligand-stabilized metal nanoparticle dispersions into narrow distributions through fine adjustments in CO2 pressure. By pressurizing an organic solution with CO2, ligand-stabilized nanoparticles were size-selectively precipitated within a novel apparatus that confined the particles to specified locations on a surface allowing their separation. Accordingly, the solvent strength of the medium was tuned through successive CO2 pressurization to provide sequential precipitation of increasingly smaller particles. A novel spiral tube apparatus was developed for separating polydisperse silver nanoparticles into different fractions of uniform sizes by regulating the CO2 pressure23 and therefore altering the liquid’s solvation of the particle ligand tails. The advantage of this apparatus is that it separates a polydisperse solution of nanoparticles into fairly monodisperse fractions in one contiguous process. This method has a number of advantages compared to traditional liquid-based size-selection

J. Phys. Chem. B, Vol. 109, No. 48, 2005 22853 methods and can be applied to a broader range of particle sizes as compared to SCF CO2 particle processing while operating at much lower pressures. This method also avoids the use of expensive and environmentally persistent fluorinated molecules commonly used in SCF processing, while simultaneously allowing for the separation of particles from a minimum amount of organic liquid. These characteristics make it a green solvent process. In this paper, a detailed study was performed to examine the factors influencing this nanoparticle size-separation process with use of CO2 as antisolvent. Variations in ligand-solvent interactions were examined to demonstrate the effects of solvent strength and thiol length on the CO2 pressure range required for particle size separation. The effect of temperature on this pressure range for precipitation was also studied. Recursive fractionation on particles collected at a given pressure range was performed to show that multiple fractionations further improve the CO2 antisolvent size-separation process, just as is observed in traditional liquid antisolvent processes. Experimental Section Materials. Silver nitrate (99.8% purity)(AgNO3) was obtained from Acros. Hydrogen tetrachloroaurate trihydrate (99.9%) (HAuCl4‚3H2O), tetraoctylammonium bromide (98%), chloroform (99.8%), sodium borohydride (99%), dodecanethiol (98%), hexanethiol (95%), octanethiol (98.5%), tetradecanethiol (98%), hexane (99%), cyclohexane (99.5%), octane (99%), and heptane (99%) were obtained from Aldrich chemical Co. Pentane (99.6%), toluene (99%), and deionized water (D-H2O) were obtained from Fisher. Ethanol (200 proof) was obtained from Florida Distillers. Carbon dioxide (SFC/SFE grade) was obtained from Airgas. All chemicals were used as supplied. Nanoparticle Synthesis. Ligand-stabilized silver and gold nanoparticles were synthesized by the two-phase arrested precipitation method as developed by Brust et al.17 In short, a solution of 0.19 g of AgNO3 in 36 mL of D-H2O was mixed with an organic solution of 2.7 g of tetraoctylammonium bromide in 24.5 mL of chloroform. The mixture was stirred for 1 h, the aqueous phase was removed, and then 240 µL of dodecanethiol was added. A solution of 0.5 g of NaBH4 in 30 mL of D-H2O was added as a reducing agent after the mixture was stirred for 5-10 min. The mixture was then stirred for 4-12 h before discarding the aqueous phase. In addition to dodecanethiol, other thiols were used to examine the effect of thiol length on the size-separation process. In each case, the mole percentage of thiol added was the same as that of the dodecanethiol described above. Gold nanoparticles were synthesized by replacing 0.19 g of AgNO3 with 0.38 g of hydrogen tetrachloroaurate trihydrate (HAuCl4‚3H2O), replacing chloroform with toluene, and adding thiol after 4-12 h of stirring. Once the thiol-coated metal particle dispersion was formed, ethanol was added as antisolvent. The dispersion of nanoparticles in the solvent/antisolvent mixture was then centrifuged (Fisher Centrific Model 228) to precipitate out the metallic nanoparticles. The particles were again washed with ethanol and centrifuged to remove any unbound ligands. This process of washing with ethanol was repeated 3 times to remove the phasetransfer catalyst. The particles were then dispersed in hexane by sonication (Fisher). The remaining dispersion of nanoparticles in hexane was used for further experimentation. UV-visible Absorbance Spectroscopy. The UV-visible absorbance spectrum of the particle dispersions in both neat solvent and the CO2-expanded solvents was measured in a highpressure view cell with a Varian 300E spectrophotometer to

22854 J. Phys. Chem. B, Vol. 109, No. 48, 2005

Anand et al.

Figure 1. Nanoparticle size-selection apparatus enclosed in a highpressure vessel shows a loading of a hexane nanoparticle dispersion. Excess hexane was loaded to saturate the high-pressure vessel with hexane vapor. A teflon fitting allows the steel rod to enter into the vessel and maintains the high-pressure seal while turning the steel rod 180°, which rotates the spiral tube with the help of the Teflon interconnect. PI and TC stand for pressure indicator and temperature controller, respectively.

monitor the precipitation of gold particles with added CO2 pressure. The cell had a stainless steel body with two O-ring sealed windows on opposite ends. The optical path length of the cell was 3 cm. A quartz cuvette of 10 mm path length was filled with 3 mL of organic solvent and 200 µL of the hexane solution of dispersed nanoparticles. A Teflon cuvette holder was then used to position the dispersion in a quartz cuvette at the centerline of the windows. The view cell was then pressurized with CO2, using an ISCO 260D syringe pump, and UV-vis absorbance spectra were collected at each operating pressure until the maximum absorbance value reached a steady value. This was an indication of an equilibrium condition being reached in terms of particle dispersion. Size-Selective Precipitation Process. The spiral tube apparatus as shown in Figure 1 was fabricated to obtain monodisperse metal nanoparticle populations23 from an initially polydisperse population through precipitation at specific locations on a surface via CO2 pressurization. This apparatus involves a 12 cm long, 2 cm diameter glass tube modified to include a concentric, spiral indentation on the surface of the tube from one end to the other. This indentation provides a 6 mm deep, 2.5 cm wide spiral channel, or groove, inside of the tube that allows a liquid droplet of nanoparticle dispersion resting within the channel to be translated from one location to another by a simple rotation of the tube while keeping the length of the tube horizontal. The spiral tube is situated within a cylindrical high-pressure stainless steel view cell equipped on one end with an O-ring sealed quartz window for observation. The other end is fitted with a Teflon tapered high-pressure fitting that allows entry of a 1/8 in. stainless steel rod attached to the spiral tube with a Teflon interconnect. This assembly allows radial rotation of the spiral tube within the high-pressure vessel by simply turning the metal rod from outside the vessel while a dynamic high-pressure seal is maintained by the Teflon fitting. The location of a liquid droplet situated in the glass tube channel (inside the tube) can then be controlled by turning the steel rod. The process was initiated by introducing 700 µL of pure hexane into the high-pressure view cell in the annular space outside of the spiral tube and allowing it to sit for at least 10 min. This hexane was introduced to saturate the system with hexane vapor prior to introducing the nanoparticle dispersion sample. This was done to minimize evaporative losses of hexane from the dispersion droplet during the separation process. Next, 250 µL of the hexane dispersion of thiol-coated metal particles was introduced into the channel of the spiral tube at the horizontal position closest to the quartz window as shown in the top image in Figure 2. The vessel was then slowly pressurized to an initial pressure of 550 psi and allowed 20 min to equilibrate at location

Figure 2. Nanoparticle size-selection spiral tube apparatus depicting recursive pressurization of organic liquid with CO2, followed by 180° tube rotations to achieve multiple size-selected populations.

A in the spiral tube. Of the overall 950 µL of hexane introduced into the 60 mL vessel (both inside and outside of the spiral tube), 15% of this hexane is dissolved into the CO2 gaseous phase at equilibrium at 500 psi and 25 °C, and this partitioning is increased to 22% of this hexane dissolved into the CO2 at the highest pressure of 700 psi as determined by phaseequilibrium calculations using the Peng Robinson equation of state. More importantly, the increased concentration of CO2 in the solvent mixture (liquid phase) decreases the overall solvent strength such that particles too large to be stabilized by the now weakened CO2/solvent mixture will precipitate during the 20 min settling time. Van der Waals forces cause the particles to adhere to the surface on which they precipitate. To separate the remaining liquid dispersion from the precipitated particles, the tube was rotated by turning the rod 180°. This rotation moves the liquid dispersion to the next location, B (180° around the tube, but further along axially), leaving behind the precipitated particles affixed to the spiral groove, at location A. The vessel was then pressurized to 600 psi with the suspension at the new location; the particles that precipitate at this pressure are, on average, smaller than those that precipitated at the lower pressure. The glass tube was then turned another 180° to take the dispersion to a new location C, leaving this second fraction of affixed particles behind in the second location B. This process was continued to acquire fractions at 625 and 650 psi at positions C and D, respectively. A final precipitation at 700 psi can induce the precipitation of the remaining particles from the hexane dispersion at location E in the spiral tube. Sample Collection. After completing the final precipitation, the vessel was depressurized. There were five particle populations in the spiral tube at locations A, B, C, D, and E. These five particle samples were recovered through redispersion in hexane, giving five different size fractions. Sample grids were made by evaporative deposition and tested for particle size distribution on a Zeiss EM 10 Transmission Electron microscope (TEM). Results and Discussion Volume Expansion of the Solution. When a given organic dispersion of nanoparticles was pressurized with CO2, the volume of the organic phase was increased by dissolution of CO2 until equilibrium was reached. This increase in volume of the organic dispersion/CO2 mixture can be characterized by the volume expansion coefficient, defined as (V - V0)/V0, where V

Size-Selective Fractionation of Metal Nanoparticles

Figure 3. Volume expansion coefficient vs system pressure for liquid hexane/CO2 mixtures pressurized with gaseous CO2 and modeled with the Peng-Robinson equation of state at 25 °C.

is the volume of the solution saturated with CO2 at a given pressure and V0 is the volume of the CO2-free solution (unpressurized). This volume expansion coefficient was estimated by using the Peng-Robinson equation of state42 and compared well to measurements made by visual observation of volume expansion in a high-pressure Jerguson sight gauge (less than 5% error between the experimental data and the equation of state in the pressure range of 500 to 700 psi). This volume expansion coefficient is necessary when interpreting UV-visible spectra to compensate for the decrease in particle concentration that accompanies an increase in solution volume with CO2 pressure. The volume expansion coefficient of hexane for a range of CO2 pressures as determined by the Peng-Robinson equation of state is shown in Figure 3 where increases in CO2 pressure significantly increase the volume expansion coefficient as a result of CO2 gas partitioning into the liquid phase. Interestingly, this dissolution of CO2 in the organic solvent also increases the density of the solvent mixture as obtained from the Peng-Robinson equation of state, indicating that while these mixtures are referred to as gas-expanded liquids, the resulting solution is a dense mixture of liquid CO2 and organic solvent. However, CO2 is a very poor solvent for the solvation of the ligand-coated particles in the organic solvent mixture. Therefore, as the percentage of CO2 increases in the expanded organic solvent, solvent-ligand interactions decrease and the dispersed particles will precipitate once a threshold solvent strength is passed. UV-Visible Absorbance Spectroscopy. An increase in CO2 pressure decreases the concentration of dispersed nanoparticles. This is due to a decrease in the solvent strength of the GEL. Here, a decrease in solvent strength means that CO2 has a very poor interaction with the n-alkyl ligand tails attached to the nanoparticles as compared to the organic solvent. So, as the concentration of CO2 in the GEL is increased, interactions between the ligand tails and the solvent are diminished such that particles are no longer stabilized and start precipitating from the solvent. The precipitation of the nanoparticles from the organic solvent depends on many factors, such as ligand type, solvent type, temperature, and metal type. The effect of each of these variables on the nanoparticle precipitation process is examined in this paper. Figure 4 presents the UV-vis spectra of gold particles synthesized by arrested precipitation and dispersed in hexane (top line). This absorption band is attributed to the absorption of Au nanoparticles dispersed in hexane and is due to the excitation of plasma resonances or interband transitions.43 The gold nanoparticles do not precipitate in the absence of CO2 even after extended periods of time. However, the intensity of the

J. Phys. Chem. B, Vol. 109, No. 48, 2005 22855

Figure 4. UV-visible absorbance spectra of gold particles dispersed in hexane/CO2 liquid mixtures at increasing CO2 pressures. The spectra were normalized to give zero absorbance at 800 nm wavelength. Decreased absorbance of gold particles after correcting for the volume expansion of hexane shows that particles are precipitating from hexane by increasing the CO2 pressure.

Figure 5. Maximum UV-visible absorbance of dodecanethiol coated silver23 and gold particles dispersed in liquid hexane/CO2 mixtures vs system pressure. Absorbance values obtained were corrected for the volume expansion of the liquid mixture.

UV absorbance band decreases when the nanoparticle dispersion was pressurized with CO2. This decrease in intensity indicates that particles begin precipitating from solution and the absorbance maximum decreases with the increase in pressure. Correspondingly, the absorbance maxima of the UV-visible spectra, after correcting for the volume expansion of the organic solvent with the addition of CO2, was plotted against the CO2 pressure as a measure of particle concentration that remains dispersed at a given CO2 pressure. Figure 5 shows a decrease of the absorbance with an increase in the pressure of CO2 and indicates that the gold nanoparticles primarily precipitate from the solution in the range of 500 to 700 psi of CO2 pressure. At pressures higher than 700 psi, complete precipitation occurs. The gold particles used in this experiment had a mean particle size of 5.0 nm and a standard deviation of 26% as shown in Table 1. McLeod et al.23 also demonstrated a dramatic decrease in the UV absorbance band for dodecanethiol stabilized silver particles dispersed in hexane with similar increases in CO2 pressure (also shown in Figure 5). The mean particle size and standard deviation for these silver particles23 was 5.5 nm and 31.9%, respectively. Interestingly, the pressure range for the precipitation and the slope of this curve was very similar for both the gold and silver nanoparticles. This is consistent with the fact that both dodecanethiol stablizied gold44 and dodecanethiol stabilized silver14 nanoparticles have Hamaker constants of 1.95 eV resulting in similar inherent van der Waals forces of attraction.

22856 J. Phys. Chem. B, Vol. 109, No. 48, 2005

Anand et al.

TABLE 1: Statistical Analysis of Particle Populations Where the Five Fractions Were Separated in a Single Experiment from the Original Populationa fraction (psi) original 0 to 550 550 to 600 600 to 625 625 to 650 650+

∆P of mean std rel std 95% fraction diameter dev dev confidence particle (psi) (nm) (nm) (%) (nm) count 0 555.0 50 25 25